New Analytical Method for the Determination of ... - ACS Publications

and 2-Methylerythritol and Its Application to Smoke and Rainwater Samples .... Partitioning of 2-Methyltetrols and Levoglucosan at an Urban Site i...
0 downloads 0 Views 308KB Size
Environ. Sci. Technol. 2005, 39, 2744-2752

New Analytical Method for the Determination of Levoglucosan, Polyhydroxy Compounds, and 2-Methylerythritol and Its Application to Smoke and Rainwater Samples G. SCHKOLNIK,† A. H. FALKOVICH,† Y . R U D I C H , * ,† W . M A E N H A U T , ‡ A N D P. ARTAXO§ Department of Environmental Sciences, Weizmann Institute, Rehovot 76100, Israel, Department of Analytical Chemistry, Institute for Nuclear Sciences, Ghent University, Proeftuinstraat 86, B-9000 Gent, Belgium, and Institute of Physics, University of Sa˜o Paulo, Rua do Mata˜o, Travessa R, 187, CEP 05508-900 Sa˜o Paulo, Brazil

Biomass burning is an important source of smoke aerosol particles, which contain water-soluble inorganic and organic species, and thus have a great potential of affecting cloud formation, precipitation, and climate on global and regional scales. In this study, we have developed a new chromatographic method for the determination of levoglucosan (a specific tracer for biomass burning particles), related polyhydroxy compounds, and 2-methylerythritol (recently identified as isoprene oxidation product in fine aerosols in the Amazon) in smoke and in rainwater samples. The new method is based on water extraction and utilizes ion-exclusion high-performance liquid chromatography (IEC-HPLC) separation and spectroscopic detection at 194 nm. The new method allows the analysis of wet samples, such as rainwater samples. In addition, aliquots of the same extracts can be used for further analyses, such as ion chromatography. The overall method uncertainty for sample analysis is 15%. The method was applied to the analysis of high-volume and size-segregated smoke samples and to rainwater samples, all collected during and following the deforestation fires season in Rondoˆ nia, Brazil. From the analysis of size-segregated samples, it is evident that levoglucosan is a primary vegetation combustion product, emitted mostly in the 0.175-1 µm size bins. Levoglucosan concentrations decrease below the detection limit at the end of the deforestation fires period, implying that it is not present in significant amounts in background Amazon forest aerosols. The ratio of daytime levoglucosan concentration to particulate matter (PM) concentration was about half the nighttime ratio. This observation is rationalized by the prevalence of flaming combustion during day as opposed to smoldering combustion during night. This work broadens the speciation possibilities * Corresponding author phone: 972-8-934-4237; fax: 972-8-9344124; e-mail: [email protected]. † Weizmann Institute. ‡ Ghent University. § University of Sa ˜ o Paulo. 2744

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 8, 2005

offered by simple HPLC and demonstrates the importance of multianalysis of several kinds of samples for a deeper understanding of biomass burning aerosols.

Introduction Aerosols from biomass burning have recently been the focus of increasing attention for their role in atmospheric chemistry and climate (1-3). Smoke particles may affect the climate directly due to their ability to absorb and scatter light (1, 4-7), indirectly by altering cloud properties, since they act as cloud condensation nuclei (CCN), and semidirectly due to their light absorption properties (6, 8-10). Clouds and precipitation on a regional scale were found to be affected by biomass burning aerosols (6, 10-13). The interaction between biomass burning aerosols and clouds may also lead to an increase in stratospheric water vapor content (12). Biomass burning aerosols consist of a complex mixture of organic and inorganic compounds. Water-soluble organic compounds (WSOC) may account for 40-74% of the smoke particles’ total carbon (14), and it is this part of the aerosol which has a significant role in determining its properties as a CCN. Levoglucosan (1,6-anhydro-β-D-glucopyranose) is a dehydrated glucose containing a ketal functional group that accounts for 2-8% of WSOC mass in Amazonian smoke samples (14). It is produced during cellulose pyrolysis (Figure 1) and emitted in fine smoke particulate matter (15-17). Thus, it is utilized as a specific tracer for emissions from vegetation combustion in atmospheric particulate matter. Combustion of non-biomass materials (i.e., fossil fuels) or biodegradation and hydrolysis of cellulose do not produce levoglucosan (18). Levoglucosan is relatively stable in the atmosphere, allowing long-range transport (19). Hence, it is highly desirable to be able to quantify levoglucosan in rainwater and smoke samples, both for source assignment purposes and as part of the WSOC characterization effort. Recently, Claeys et al. (20, 21) have identified 2-methylerythritol and 2-methylthreitol in background aerosol particles from Amazonia. It was suggested that these diastereomeric five-carbon compounds are formed through the liquid-phase oxidation of isoprene by hydrogen peroxide and account for ∼2% of the fine ( 300 °C (17). GC/MS and a method based on HNMR detection, which as yet has not been validated with respect to levoglucosan (24). Here, we describe the development of a new method which is based on ion-exclusion chromatography (IEC) and allows detection and quantification of levoglucosan and other polyhydroxy compounds in rainwater and water-extracted smoke samples. This is a relatively simple method which is based on water extraction of these species. Aliquots of the same sample extract may be used for further ion chromatographic analyses. The application of the new method to the analysis of smoke particles and to rainwater samples from Amazonia is reported, and conclusions are drawn as to the use of levoglucosan as a biomass burning tracer.

Experimental Section Materials. Perchloric acid solution (10 mM) was used as an eluent for the liquid chromatography. The eluent was prepared from perchloric acid 70% (AnalaR, BDH Laboratory Supplies, Pool, England). Standards used were 1,6-anhydroβ-D-glucopyranose (levoglucosan) g98.0% and D(+)-arabitol g99% (Fluka Chemie, GmbH), meso-erythritol (Aldrich Chemical Co., Wisconsin, United States), D(-)-mannitol 98101% (Riedel-de Hae¨n, Sigma Aldrich GmbH), 1,6-β-Dmannopyranose (mannosan) ∼98% (Sigma, MO), and Dglucose (AnalaR BDH Laboratory Supplies, Pool, England). A mixture of 2-methylthreitol and 2-methylerythritol, synthesized by the University of Antwerp (20), was used as a standard. All water used had resistivity of 18.2 ( 0.1 MΩ-cm. Sampling. Smoke samples were collected in the Amazon during the deforestation fires season. Sampling took place at the Fazenda Nossa Senhora Aparecida pasture site, in Rondoˆnia, Brazil, September and October 2002, in the framework of the LBA-SMOCC (The Large Scale BiosphereAtmosphere Experiment in AmazoniasSmoke Aerosols, Clouds, Rainfall, and Climate: Aerosols from Biomass Burning Perturb Global and Regional Climate) field campaign. All samples were collected for 12 h for high loading and 24 h for cleaner conditions. High-volume (HiVol) samples were collected by the Institute for Nuclear Sciences, Ghent University (UGent), Belgium, using a high-volume dichotomous sampler with two size fractions, out of which the fine size fraction (500 °C prior to sampling to eliminate organic contamination. During sampling, a back filter was placed behind the sample filter to assess sampling artifacts in aerosols collection (25), such as gas-phase species that adsorb on both filters. Samples were stored at -25 °C in prebaked aluminum foil envelopes. Some additional samples used only for method development were collected in Rehovot, Israel, on May 1011, 2001, during a national bonfire festival in Israel. They were collected on Gelman Quartz Microfiber (QF) filters (Gelman, Pall Corporation, NY) using a commercial highvolume sampler. Size-resolved samples were collected by the Institute of Physics, University of Sa˜o Paulo (IFUSP), Sa˜o Paulo, Brazil, using a microorifice uniform deposit impactor (MOUDI) (model 110, MSP corporation, Minneapolis, MN), on Nuclepore Polycarbonate filters (Whatman, New Jersey) of 47-mm diameter, at a flow rate of 25 L/min. Aerosols of aerodynamic diameter less than 18 µm were separated into nine stages with calibrated aerodynamic cutoffs of 18, 10.0, 3.2, 1.8, 1.0, 0.56, 0.33, 0.175, and 0.093 µm. The inlet stage was not analyzed in this study. The loaded filters were placed immediately in clean plastic Petrislides and stored in the dark at -18 °C until analysis. Field blanks were obtained using the same loading and unloading procedure as for normal filters but with a sampling time of only 15 s. Rainwater sampling was performed from 12 September to 10 November using a wet-only Aerochem-metrics sampler. Rain was collected in high-density polyethylene bottles on an event basis and stored immediately after collection. Thymol was added to the bottles prior to sampling to preserve organic species from bacterial activity. The samples were stored in the darkness under refrigeration until analysis. Sample Preparation. Extraction. Each HiVol sample (approximately 15 cm2) was extracted twice into 5.0 mL of water by short vortex agitation followed by 15 min of gentle shaking. The combined extract was centrifuged for 5 min and filtered through a GHP Acrodisk syringe filter (25 mm, 0.45-µm pore size, Gelman, Pall Corporation, NY), which was previously washed with 10 mL water. MOUDI samples were extracted in the same manner into 4.0 mL of water. These were filtered using a GHP Acrodisk syringe filter (13 mm, 0.45-µm pore size). It has been validated that further extraction was not needed. Rainwater samples were concentrated by a factor of 3-12 under nitrogen flow, at 30 °C, and were then directly injected. Solid-Phase Extraction (SPE). Polyacidic humic-like substances (HULIS) are present in smoke samples from Amazon forest fires (26). Since these compounds are negatively charged, they are eluted in the void volume of the ionexclusion column (see below), and due to their strong light absorption, they form a huge unresolved peak in the beginning of the chromatogram. The big tail of this peak interferes greatly with the detection of the neutral polyhydroxy compounds, and to avoid that, the polyacids must be selectively removed. Therefore, each sample (both rainwater and filter extracts) was passed through an Accell QMA SPE anion exchange cartridge (Waters, MA), which retains dissociated acids, such as the HULIS, but not neutral compounds, such as polyhydroxy compounds. This was done according to sample size: Plus cartridges (internal volume 0.8 mL) were used for the larger HiVol and rain samples, while Light cartridges (internal volume 0.4 mL) were used for the smaller MOUDI samples. Each Plus cartridge was prewashed with 6 mL of water, and then 4.0 mL of the sample was passed through it. A levoglucosan retention test was conducted for the SPE cartridges. It was found that the first 0.5 mL eluting from Plus SPE cartridges do not contain levoglucosan (due to the cartridge’s dead volume). Hence, VOL. 39, NO. 8, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

2745

FIGURE 2. Absorption spectra of glucose, mannitol, arabitol, erythritol, and levoglucosan (polyhydroxy compounds) and of the eluent, perchloric acid 10 mM. As one can see from the figure, within our detector’s range the polyhydroxy compounds show high absorption at 194 nm, where the eluent has insignificant absorption. it was discarded and the next 3.5 mL plus an additional 1 mL of water that was passed through the cartridge was collected for each sample. Each Light cartridge was prewashed with 3.5 mL of water, and then 1.5-2.3 mL of the sample was passed through it. Since the dead volume of these Light SPE cartridges is very small, as was verified in a levoglucosan retention test, the entire sample was collected, along with an additional 0.5 mL of water. IEC-HPLC-PDA Analysis. Chromatographic separation of polyhydroxy compounds (anhydrosugars, sugar alcohols, polyols, etc.) was achieved by ion-exclusion chromatography (IEC) in a poly(styrene-divinylbenzene) sulfonate (H+) cationexchange resin column (Dionex, IEC-AS1). Unlike ionexchange chromatography, in IEC the charge of the functional groups on the ion-exchange resin has the same sign as that of the analytes. In other words, weak acids, alcohols, and carbohydrates are separated on a cation-exchange resin in the H+ form, using an acidic eluent, by a hydrophobic adsorption-retention mechanism (27). In IEC, strong acids, being highly ionized, pass quickly through the column, while nonionic compounds can enter the resin network and elute in order of decreasing acidity (27, 28). The development of the IEC separation procedure applied in this study is based on the Dionex application note for separation of aliphatic alcohols on an ICE-AS1 column, utilizing 50 mM perchloric acid as eluent at 0.8 mL/min flow rate (40). According to the IEC principles, aliphatic alcohols which contain several hydroxyl groups are eluted in order of decreasing number of hydroxyl groups (i.e., decreasing acidity). A Varian ProStar 230I HPLC pump, a Varian ProStar 410 autosampler, and a Varian ProStar 330 photodiode array (PDA) detector were used. The aim of the procedure development was to optimize separation between glucose, mannitol, arabitol, erythritol, 2-methylerythritol, 2-methylthreitol, and levoglucosan (polyhydroxy compounds hereafter). The retention of the polyhydroxy compounds was constant for eluent (perchloric acid) concentrations ranging between 10 and 100 mM. Therefore, elution was performed with 10 mM perchloric acid at 1 mL/min flow rate. Two columns were sequentially connected to improve separation. To minimize peak widths and stabilize the baseline, the columns were thermostated at 26 ( 0.5 °C. At 194 nm, the polyhydroxy compounds have sufficient absorption, while the eluent’s absorption is very low (Figure 2). 2746

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 39, NO. 8, 2005

FIGURE 3. Chromatogram of a high-volume PM2.5 smoke sample, collected in Rondoˆ nia, Brazil, during the deforestation fire season. The peaks: (1) the remains of the HULIS front peak, (2) 2-methylerythritol, (3) levoglucosan, and (4) mannosan, a stereoisomer of levoglucosan, that appears as a shoulder on the levoglucosan peak.

FIGURE 4. Three chromatograms demonstrating standard separation obtained by the method. Solid line: mannitol, arabitol, erythritol, and levoglucosan (in that order) are well separated (chromatographic resolution, R > 3.0). Dashed line: mannosan, a stereoisomer of levoglucosan, which can be separated from levoglucosan by multiple-Gaussian peaks fit analysis. Dotted line: 2-methylerythritol (left) is separated from levoglucosan (R ) 1.7), while 2-methylthreitol (right) is not separated from levoglucosan. The coelution of 2-methylthreitol with levoglucosan must be considered in calculations. Integration of the chromatograms was performed using the Origin graphics program by multiple-Gaussian-fit analysis after baseline subtraction. A typical smoke sample chromatogram obtained by the method is shown in Figure 3. As can be seen in the figure, the front HULIS peak (peak 1) is large even after SPE treatment. Without the use of SPE, this peak covers most of the peaks seen in the chromatogram. Validation. Chromatographic Separation and Levoglucosan Determination. Glucose, mannitol, arabitol, erythritol, and levoglucosan are well separated (Figure 4) with R > 3.0, (R is the chromatographic resolution, which is defined as the ratio between twice the distance between two peaks and the sum of their widths at half-height) (29). 2-Methylerythritol is separated from levoglucosan with R ) 1.7. However, levoglucosan coelutes with 2-methylthreitol, and this must be taken into account in the quantitative analysis. Mannosan,

a stereoisomer of levoglucosan, appears as a shoulder on the levoglucosan peak (Figure 3) and is resolved using the multiple-Gaussian peaks fit analysis. Coelution Correction. The University of Antwerp team has shown that the ratio between the two 2-methyltetrol stereoisomers in smoke samples is almost constant (41). The ratio deduced from the analysis of HiVol samples collected in parallel to the ones reported in this paper, during the SMOCC campaign, is 2-methylthreitol/2-methylerythritol ) 0.34 ( 0.07. Therefore, to compensate for the coelution of levoglucosan with 2-methylthrietol, we have subtracted 0.34 of the 2-methylerythritol peak area from the levoglucosan peak area. This correction is typically 4% of the levoglucosan signal. Linearity, Limit of Detection (LOD), Limit of Quantification (LOQ), and Blank Control. The linear range of levoglucosan detection is between 2.5 µg/mL and 25 µg/mL (R2 ) 0.9993 for the regression in the linear range). The limit of detection (LOD) is 0.5 µg/mL with signal-to-noise ratio (S/ N) ) 3 for five repetitions. The limit of quantification (LOQ) is 1 µg/mL, with S/N ) 4.5 and %RSD ) 9% for 11 repetitions. Blanks. For all the steps of the procedure, blank measurements were conducted. All blanks, including lab blanks, sampling field blanks, and HiVol sampler back filters, showed no levoglucosan. Recovery. Recovery Test and Levoglucosan Stability. To determine the efficiency of the extraction from smoke samples, the recovery of the entire procedure was verified by spiking quartz fiber filters with 10 µg levoglucosan standard and then subjecting them to the described procedure. On the basis of eight replicates, the recovery obtained was 95 ( 3%. Since we utilize water extraction, the same extract used for levoglucosan determination was also used for ion chromatographic analysis. For this reason, samples were not always analyzed immediately after extraction, and the extracts were stored in refrigeration (2-4 °C) for 1-8 weeks. Levoglucosan standards were stored under the same conditions to test levoglucosan stability. No significant loss of levoglucosan during storage was observed. Precision. The maximum RSD obtained for repeated injections of the same standard solution of levoglucosan (concentration 2 µg/mL) was 6% for five injections. To determine the precision obtained for several extractions of the same sample filter, we used a sample collected in Israel during a national bonfire festival on May 10-11, 2001. The sample was divided into six parts and the same procedure was repeated for each part. The precision obtained was RSD ) 3.3%. Since the deviation between injections of the same extract is 9%, larger than the deviation between fractions of the same filter, we take the overall sample RSD to be 9%. Overall Method Uncertainty. The method uncertainty for detection of a standard is 6%, which is the worst-case precision of standard detection. In real smoke samples, additional sources for uncertainty must be considered, including (1) the analytical procedure uncertainty, including the injection precision of both standard and sample solutions and the recovery uncertainty, and (2) the uncertainty of the correction needed due to the coelution of levoglucosan with 2-methylthreitol. The method’s uncertainty in levoglucosan concentration, prior to the coelution correction, was 11%. The coelution correction adds 4%. Thus, the overall uncertainty of levoglucosan determination in smoke samples is 15% for concentrations greater than LOQ. For concentrations that lie between LOQ and LOD, an uncertainty of 23% is assigned. All method validation parameters are summarized in Table 1. Applicability of the Method. The new method employs shorter sample preparation times and is simpler than the existing derivatization-GC/MS method. Since it utilizes water

TABLE 1. Summary of the Method’s Validation Parameters linearity range limit of detection limit of quantification recovery precision (standard) precision (smoke sample) overall method uncertainty (for concentrations > LOQ)

2.5-25 µg/mL 0.5 µg/mL 1 µg/mL 95 ( 3% 6% 9% 15%

FIGURE 5. A comparison between levoglucosan values obtained using the method described in this paper (y-axis), and with values obtained using the silylation-GC/MS method (Claeys, unpublished data, x-axis) for samples collected in parallel. The correlation slope is 1.02 ( 0.05, and the intercept is at -0.14 ( 0.09. R2 ) 0.9803. extraction, it retrieves the fraction pertinent to CCN formation and allows the analysis of wet samples such as rainwater and cloud droplets. Ion chromatography can be further carried out on the same extracts. The main drawback of this method in comparison to the GC/MS method is its lower sensitivity. The GC/MS method can detect levoglucosan in background aerosols while this method is as yet unable to. Sensitivity may be increased by further concentrating water extracts and by using SPE cartridges with smaller internal volumes. Another issue is the coelution of levoglucosan with other species, which was overcome in this work by the integration technique and by a priori knowledge about the ratio between the 2-methyltetrol diastrereomers. Further investigation of this ratio in different conditions is therefore desirable to reduce the portion of method uncertainty caused by the uncertainty in this ratio. Since these compounds have been just recently discovered in aerosols (20), it is to be expected that such information will soon follow. Finally, due to both separation and sensitivity issues, this method cannot be used to quantify mannosan and galactosan, levoglucosan’s stereoisomers, emitted in much smaller quantities than levoglucosan (30) and glucose (due to coelution with a component from the SPE cartridges), a biogenically emitted compound. In view of the merits and shortcomings of the new method, it is suggested that the derivatization-GC/MS method is most suitable for individual quantification of each polyhydroxy compound in a small amount of samples, which may include either smoke or background aerosols. In contrast, the new IEC-HPLC method is most suitable for quantifying levoglucosan or 2-methylerythritol, combined with full ion analysis of a large amount of samples collected under smoky or semismoky conditions or of wet samples such as rainwater and cloud droplets. In the next section, such a multisample integrated analysis is demonstrated. Method Application: Results and Discussion. HighVolume Samples. During the LBA-SMOCC field campaign, VOL. 39, NO. 8, 2005 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

2747

TABLE 2. Levoglucosan Concentrations in High-Volume PM2.5 Smoke Samples Collected during the SMOCC Campaignb sample name R2HIV03D R2HIV03N R2HIV04N R2HIV05D R2HIV05N R2HIV06D R2HIV06N R2HIV07D R2HIV07N R2HIV09D R2HIV09N R2HIV10D R2HIV10N R2HIV11D R2HIV11N R2HIV14D R2HIV14N R2HIV15D R2HIV15N R2HIV16N R2HIV17D R2HIV17N R2HIV18D R2HIV19D R2HIV19N R2HIV20N R2HIV21 R2HIV24D R2HIV24N R2HIV26D R2HIV26N R2HIV29D R2HIIV29N R2HIV30 R2HIV32 R2HIV35D R2HIV35N R2HIV39 R2HIV40 R2HIV41 R2HIV43Na R2HIV44Da R2HIV45Na R2HIV46Da R2HIV48Da R2HIV49Na R2HIV54

start time GMT (2002) Sept 11 Sept 11 Sept 12 Sept 13 Sept 13 Sept 14 Sept 14 Sept 15 Sept 15 Sept 17 Sept 17 Sept 18 Sept 18 Sept 19 Sept 19 Sept 22 Sept 22 Sept 23 Sept 23 Sept 24 Sept 25 Sept 25 Sept 26 Sept 27 Sept 27 Sept 28 Sept 29 Oct 2 Oct 2 Oct 4 Oct 4 Oct 7 Oct 7 Oct 8 Oct 10 Oct 13 Oct 13 Oct 17 Oct 18 Oct 19 Oct 21 Oct 23 Oct 24 Oct 26 Oct 29 Oct 30 Nov 10

12:38 22:15 22:22 12:29 22:21 12:14 22:28 11:27 22:30 11:45 22:45 12:45 22:50 12:40 22:30 11:45 22:30 11:45 22:45 23:00 13:00 23:00 11:50 12:00 22:30 23:30 12:15 13:15 22:48 12:00 22:45 12:00 22:30 14:15 12:00 12:15 22:15 12:15 12:00 12:00 23:30 13:00 23:00 12:00 15:30 23:30 12:30

end time GMT (2002) Sept 11 Sept 12 Sept 13 Sept 13 Sept 14 Sept 14 Sept 15 Sept 15 Sept 16 Sept 17 Sept 18 Sept 18 Sept 19 Sept 19 Sept 20 Sept 22 Sept 23 Sept 23 Sept 24 Sept 25 Sept 25 Sept 26 Sept 26 Sept 27 Sept 28 Sept 29 Sept 30 Oct 2 Oct 3 Oct 4 Oct 5 Oct 7 Oct 8 Oct 9 Oct 11 Oct 13 Oct 14 Oct 18 Oct 19 Oct 20 Oct 23 Oct 24 Oct 26 Oct 27 Oct 30 Nov 1 Dec 11

21:42 11:40 11:36 21:45 11:37 21:16 10:51 21:45 10:58 21:45 11:00 21:40 11:00 21:45 11:00 21:45 11:00 22:00 11:00 11:00 22:45 11:00 21:45 21:45 11:00 11:00 11:00 21:45 11:00 21:45 11:00 21:45 9:10 11:00 11:00 21:45 11:00 11:00 11:00 10:40 11:00 22:00 11:00 22:00 22:00 11:00 12:50

levoglucosan [µg/m3]

% levoglucosan carbon of WSOC

PM2 [µg/m3]

K+ [µg/m3]

0.08 0.15 1.14 0.09 2.65 0.81 0.60 0.34 0.53 1.34 3.37 0.67 3.81 0.93 5.80 0.86 0.67 1.03 0.98 1.13 1.98 1.70 0.34 0.17 0.20 0.22 0.11 0.23 0.94 0.80 4.17 0.21 0.32 0.11 0.09 0.03 0.33 0.54 0.53 0.26 0.17